Research On Image Stitching Technology Based On Image Matching

Image mosaic is a digital image processing technology that stitches a series of images with overlapping fields in the same scenes and builds a panoramic image which contains high resolution and wide viewing angles. It requires the panorama is as close as possible to the original image and the distortion is as small as possible, and no visible boundary on panorama.With the development of digital image technology, high resolution, wide view image is more and more important in a number of important areas, the panorama image stitching technology can get high-resolution, wide field of view. It can supply solutions to these important areas.Therefore, this paper research the image mosaic technology based on the static image.In this paper, firstly, we analyze and study the method and theory of image mosaic through the description of the steps of image mosaic. Secondly, on the basis of the comparative analysis of the SIFT algorithm and the SURF algorithm, the paper focuses on the graph based on SURF feature.The shortcomings of the traditional SIFT algorithm is that the computation speed of feature extraction and matching is slow, while SURF solves this problem,and has high robustness. In this paper, the SURF algorithm is improved by the research of every step of the SURF algorithm. For the feature points extracted by SURF algorithm can be classified with Laplacian identifier according to the ratio of the nearest neighbor and the adjacent distance. In the calculation of the matrix changes, because the traditional RANSAC algorithm need multiple iteration, and the calculated data is not accurate enough, in order to solve this problem, in this paper, we select the similarity value of matching pair to ensure the accuracy of registration. Then, in image fusion, in order to remove the stitching traces may produce in the process of stitching, we improved the traditional method gradually in and out.Using the Gauss model to calculate the weight to replace the traditional calculation method, to improve the effectiveness of remove the seam. This paper calculates the image region of interest first, the Gauss model is then used to calculate the weights. Finally, the weights of region of interest(ROI) were fused to remove the seam. Experiments show that thealgorithm proposed in this paper is effective and feasible.